Issue running Biowdl Germline-DNA v4.1.0: unusable runtime struct initializer

Full error:

workflow Germline (germline.wdl Ln 34 Col 1) failed :: dir: "/exports/sasc/rhpvorderman/testminiwdl/germline-dna/20220802_153134_Germline", error: "EvalError", message: "unusable runtime struct initializer (member type mismatch, lacking required member, or extra member)", node: "call-sampleWorkflow-0", pos: {"source": "/exports/sasc/rhpvorderman/testminiwdl/germline-dna/germline.wdl", "line": 119, "column": 26}

The WDL file and its imports can be downloaded here: https://github.com/biowdl/germline-DNA/releases/tag/v4.1.0 (germline_v4.1.0.wdl and its accompanying imports zip).

It works fine in Cromwell. There are some optional members in the structs that we use. Apparently miniwdl cannot handle that properly. Unfortunately I cannot get a more descriptive error message regarding which member.

[rhpvorderman]

Hi, sorry to reopen this. I have a similar error now with the gender member of the struct. That is not defined in the JSON. As a result MiniWDL crashes on gender = select_first([sample.gender, "unknown"]), in the wdl file. So miniwdl does not complain about undefined Struct members in the JSON if they are optional, but also does not add them to the struct and set them to None. The latter option is preferred to complaining in my opinion.

EDIT: I did some further debugging. It seems it goes wrong in the value.from_json function. Where the struct for the SampleConfig is correctly loaded yet the nested struct types are inferred from json (_infer_from_json) rather than serialized to their native struct types.

EDIT2: Have been able to fix it in the wdl with the following code:

    scatter (sample_obj in sampleConfig.samples) {
        Sample sample = sample_obj

Where previously scatter (sample in sampleConfig.samples) was used. So I made the coercion explicit. However, since the struct fields have explicit Sample types, the coercion should also happen in the simpler scatter example. I am also happy to report the workaround does not break cromwell, so at least I have a good workaround now for this issue.

[rhpvorderman]

Yes #581 Solved this particular issue on the germline-dna pipeline. Great, thanks!